Brief in progress — review the script and demo before recording.
The previous lesson looked at multi-agent systems from a technical perspective. This video shifts to practical workflows. We can draw on Goose use cases again, but, where possible, we should also mention other solutions so the video feels broadly applicable rather than promotional.
- Setup: introduce patterns such as hierarchies, blackboards, networks, and graphs. We do not need to walk through their implementations; the goal is to show that these patterns exist and when they can be useful.
- Knowledge: agents need both narrow and broad context to understand a task and make good decisions. We could show agents working from a Kanban board or using a similar approach, but the main takeaway is that the knowledge system needs to fit the project and the team’s preferences.
- Communication: multi-agent workflows often run into the limits of human attention: agents write faster than humans can read. We need to show how to organize work between agents and humans, while encouraging students to start small and gradually improve the workflow together with the model.
- Improvements: agents can scan conversation history, review recent work, and use that context to suggest improvements. We could demonstrate a workflow built around this feedback loop.
We may want to showcase scenarios such as:
- A look at how Goose is developed. Since this video focuses on multi-agent systems, Angie, choose the format that best fits your experience. Showing work on a real, open-source project would be especially valuable.
- Alternatively, a few simplified demos that illustrate the most important concepts above.